Luminescent Tower

Risk lived in a hundred spreadsheets

Before this project, risk tracking happened wherever each team could manage it — spreadsheets, wiki pages, email threads. There was no shared definition of what a "risk" was, no consistent severity scale, and no way for leadership to see aggregate exposure across orgs.

That created three compounding problems: risks were logged inconsistently, ownership was ambiguous, and the same issue got reported multiple times by different teams. Leadership was making decisions on an incomplete picture.

Design question

How do we create one system that is structured enough to aggregate risk data reliably, but flexible enough that a dozen very different teams will actually adopt it?

How I approached it

Talking to the people who log risk

I interviewed 14 people across three roles — risk owners who log issues, managers who triage them, and leaders who report upward. I also audited 20+ existing spreadsheets to understand what people were actually tracking versus what the official process asked for.

What I learned

• Logging friction was the real blocker. People skipped the official process because it took 20+ minutes per entry.
• Nobody trusted the data. Because severity was subjective, teams discounted each other's ratings.
• Triage was invisible work. Managers spent hours de-duplicating and re-routing risks by hand.
• Leaders wanted trends, not lists. A flat table of open risks did not answer "are we getting better or worse?"

Structure has to be earned, not enforced

The instinct was to mandate a rigid taxonomy. But research showed that every previous attempt at enforcement had failed — people simply worked around it. The insight that reframed the project: if logging a risk is faster than not logging it, structure becomes a byproduct rather than a burden.

So instead of a long compliance form, I designed a progressive flow — capture the minimum viable risk in under a minute, then enrich it later as triage progresses. Structure emerges through the workflow rather than being demanded upfront.

From sketches to structure

I explored three information architectures: a linear queue, a severity-first matrix, and an ownership-based dashboard. Testing showed the dashboard model matched how managers actually think — they scan by ownership first, then drill into severity.

Testing what we chose

I ran moderated usability sessions with 8 participants on an interactive prototype. Two changes came directly out of those sessions: severity became a guided set of questions rather than a free choice, and duplicate detection surfaced inline while typing instead of after submission.

One system, three views

Quick capture

A single-screen entry flow that asks only what's needed to route a risk. Inline duplicate detection catches overlaps as the user types, cutting redundant entries significantly.

Guided severity

Rather than asking users to pick a number, the system asks about impact and likelihood in plain language and derives severity consistently. This made ratings comparable across teams for the first time.

Ownership dashboard

Managers see their team's exposure grouped by owner with trend indicators over time. Leadership gets the same data rolled up, so conversations start from a shared source of truth.

What changed

• Time to log and route a new risk dropped by roughly 60%.
• 12 teams onboarded in the first quarter, without a compliance mandate.
• Severity ratings became comparable across orgs, enabling the first aggregate risk view at this scope.
• The component patterns built here were adopted by two adjacent governance tools.

Note: metrics shown are directional and rounded. Detailed figures are confidential.

What I'd do differently

I under-invested in the triage experience early on. I focused on capture because that was the loudest complaint, but managers turned out to be the group whose time was most wasted. If I ran this again, I'd map the full lifecycle before choosing where to start.

The lesson that stuck: adoption is a design problem, not a rollout problem. Every place we replaced a mandate with a genuinely faster path, usage followed on its own.